TL071 offset null... what's this circuit doing with it?

[aion]

So I traced the 2014 version of the Hotcake, the one with an internal switch for what's called "Bluesberry" mode.



The schematic is still missing some values, but I immediately noticed some weirdness around pin 5 of the IC, the offset null pin, and wanted to ask what was happening here - both what function it serves and what's happening when SW1 is flipped. I've never seen anything quite like this before.

There is always the possibility that I got something wrong in the trace, but I did go over it a few times and am pretty confident in the bones of it. But I can re-check if anything in particular looks suspect.

[antonis]

Maybe we have to refer to TL071 internal architecture diagram, 'cause Q1/Z1 & relevant circuitry doesn't pose weird analysis difficulty but "open" pin 1 DOES a lot..

[Rob Strand]

IIRC the idea is it allows the positive clipping point to be set using the voltage on the base of the PNP transistor.

The way it works is when the output rises on the positive swing it lifts the emitter of the PNP transistor.  When the output reaches 0.6V above the voltage set on the base a collector current flows into pin 5.   The base voltage is set at Vcc-2.7 or Vcc (disable PNP ckt).

To understand how and why driving current into pin 5 does actually limits the output swing below the positive power rail you need to look at the internal schematic of the TL071.

One thing this method does is clip the opamp *without* the opamp's output stage saturating against the positive supply rail.   That probably leads to much cleaner clipping.

Another point is feeding current into pin 5 might (need to check TL071 insides to confirm) act to limit the output swing of the opamp via feedback,  not by clipping the opamp.   This subtle difference also means the clipping will be cleaner as the opamp is still operating linearly.
--------
Edit:  An interesting point is, if the collector of the PNP was wired back to pin 2 instead of pin 5 would it sound the same?     That would allow opamps to be used which don't have the offset pin.

[aion]

Thank you! This circuit is notable (or notorious depending who you ask) for its opamp clipping, not a diode in sight, so good to have confirmation that this arrangement does something that would directly impact the character of the clipping.

Another question: what's the 2.7V zener doing? I see that one side of the switch bypasses the zener and connects the bias supply directly to VA, but not sure exactly what it does to have the zener connected with its cathode to a higher voltage.

[antonis]

As I see it, with SW1 at 2-1 position , Q1 does absolutely nothing..
(its Base is at 9V where its Emitter is "ideally" at the same voltage - Q1 OFF..)

With SW1 at 2-3 position, I realize that Zener is useless..!! 

P.S.
I take your word for accurate circuitry tracing, but..

[aion]

With SW1 at 2-3 position, I realize that Zener is useless..!! 

P.S.
I take your word for accurate circuiy tracing, but..

I'll double check the orientation. It does seem that it would make more sense going the other direction, dropping 2.7V from the supply, and then the R11 resistor adjusting the reference voltage ratio at the same time. In the SW1 2-1 position, if Q1 is disabled and R11 is out of play, it's very close to the 2003 version of the Hotcake (based on Analogguru's trace).

[antonis]

It's late here (and a snail-raki meal isn't the best advisor..) but I can't get the reason for having various voltage sources "fighting" when SW1 at pos 2-3..

[Rob Strand]

The purpose of the zener is give to a reference voltage which is 2.7V below the positive rail.  When the base of Q1 is connected to +V, Q1 is disabled and the opamp clips naturally.  When the base of Q1 connected to the zener the opamp will clip cleanly like I described before at a voltage 0.6V above the voltage on the zener - ie at 2.7 - 0.6 = 2.1V below the positive rail.   The zener voltage is probably purposefully chosen so the Q1 circuit can control the clipping *without* the IC clipping.

The other action of the switch is to shift the DC bias of the opamp.   So one action is to change the positive clipping point and the other is to tweak the bias voltage.   

There's no theory which can be applied to determine how the switch changes the bias.  The biasing is set by the designer.    The circuit is what it is.   Any use of theory could override the designer's intent.   

If you have a unit then it would be possible to:
- measure the voltages on the various resistors.
- measure the DC bias voltage on the opamp.

From the way the bias moves up and down and also the actual voltages you can get an idea how the switch should be wired.   You then wire it up like that on a prototype and check that the DC voltage at the various points in the circuit match the real unit.

Finally put the prototype side-by-side with the real unit and do an AB comparison of the sound.

FWIW, the Hotcake schematic with the transistor was around on the internet.  There might be a couple of schematics.  I can't remember if it had a zener.

[aion]

The purpose of the zener is give to a reference voltage which is 2.7V below the positive rail.  When the base of Q1 is connected to +V, Q1 is disabled and the opamp clips naturally.  When the base of Q1 connected to the zener the opamp will clip cleanly like I described before at a voltage 0.6V above the voltage on the zener - ie at 2.7 - 0.6 = 2.1V below the positive rail.   The zener voltage is probably purposefully chosen so the Q1 circuit can control the clipping *without* the IC clipping.

Thanks, this is helpful! (BTW - I did confirm the orientation of the zener and it definitely goes cathode to VA, as drawn.)

If you have a unit then it would be possible to:
- measure the voltages on the various resistors.
- measure the DC bias voltage on the opamp.

From the way the bias moves up and down and also the actual voltages you can get an idea how the switch should be wired.   You then wire it up like that on a prototype and check that the DC voltage at the various points in the circuit match the real unit.

Finally put the prototype side-by-side with the real unit and do an AB comparison of the sound.

Yep... these things are gooped to infinity with a rather persistent caulk-like substance so it didn't survive the trace unfortunately. I've got good photos of the degooping process so I can cross-check and reconstruct as needed, but it's not currently in a measurable state

FWIW, the Hotcake schematic with the transistor was around on the internet.  There might be a couple of schematics.  I can't remember if it had a zener.

I'd love to see this if you have any recollection of where you might have come across it! I saw one schematic that had a diode across the inverting and non-inverting inputs, and one from inverting to ground, but there wasn't any info on where the schematic came from - either a real trace or just a DIY mod someone came up with.

[Rob Strand]

Yep... these things are gooped to infinity with a rather persistent caulk-like substance so it didn't survive the trace unfortunately. I've got good photos of the degooping process so I can cross-check and reconstruct as needed, but it's not currently in a measurable state

Oh well, that's life I guess.

If you are confident the switch wiring hasn't been damaged or confused in the degooping process then it's really only the value of the 220k that's up for grabs (if at all).    That's means what you have drawn is pretty much it.      The way you have drawn the schematic doesn't look like anything is obviously wrong.

If you aren't confident of the switch wires then that's more trouble.

With the switch in "standard" mode, where the base connects to +V, the 82k also connects to +V.     That configuration matches the 2003 schematic, so it should give some confidence of being correct.   It's not unreasonable that "standard" mode is an option.

If we say standard mode is an option then I see only two options that make sense:  the connection you have on the schematic, and another one where pins 1 and 2 of the switch are swapped.   In that second option the 220k would connect to +V.   

Given the zener is operating at low current it might only have a drop of 2.2V instead of 2.7V

For the way you have the schematics now the opamp will bias in the region of 3.46V to 3.74V and the clip point will be 6.9V to 7.4V.    So that means the opamp is biased to about 0.50 times the clip-point.    Exactly half way.   Looks fairly deliberate.

For the alternative  connection the bias point be higher,   around 4.40V to 4.63V and the clip point will still be 6.9V to 7.4V.  So a that means the opamp is biased about 0.63 times the clip point.     Not an unreasonable value.

The difference should be audible on a side-by side comparison.

I'd love to see this if you have any recollection of where you might have come across it! I saw one schematic that had a diode across the inverting and non-inverting inputs, and one from inverting to ground, but there wasn't any info on where the schematic came from - either a real trace or just a DIY mod someone came up with.

It's hard to know.    I did have a look around yesterday but I couldnt find anything.   The internet used to have a lot of smaller collections of schematics but they are gradually disappearing and there's more a culture of posting vero layouts without schematics now.

[aion]

Just as a followup to this - I posted a walkthrough of the Hot Cake trace for anyone interested:

https://aionfx.com/news/tracing-journal-crowther-hot-cake/

...including a reference back to this thread. I did build up a prototype based on the trace and it sounds great, so I think all the conjecture here (especially Rob!) was spot on.

[Vivek]

Do all these high falutin ways of clipping actually lead to something radically different than a set of diodes making up a DFG ?

[Rob Strand]

Do all these high falutin ways of clipping actually lead to something radically different than a set of diodes making up a DFG ?

The main goal of the Hot Cake ckt is to stop the opamp clipping.   The signal is clipped by feedback not by the opamp output saturating.     It's not a new idea even the basic feedback diodes produce a clipped output without the opamp clipping because the feedback is at work.    When the output of an opamp clips it has an impact on the sound, which depends on the opamp.   The main characteristic is the clipping is hard with a sharp knee.

Another difference is the clip levels of a saturating opamp are determined by the supply voltage whereas the clip levels of a diode clipper are determined by the diode voltage.       In the feedback case, the clip point doesn't need to be the whole supply voltage you can use dividers to have clip point some fraction of the supply, or in the case of the Hotcake a fixed *DC* voltage set by the zener.   This scheme means the non-linearity does have a programmed knee which is not just coming from diode characteristics.   The knee is therefore quite sharp.  It isn't perfectly sharp as diodes or transistors are required for the circuit to work and that creates a transition region.


There's many types of diode function generators (DFG).  Some the thresholds are set by only the diodes others are set by the power supply and the diodes.   Also some diode function generators are open loop.   Only for the cases where feedback is present and the power supply sets the knee point can we say it is similar to the Hot Cake.

If you look at this circuit it uses diodes but the clip points are programmed from the power rails:


At the bottom of reply #2 I mentioned modifying Hot Cake to connect the transistor differently so you can get the same effect without requiring an offset pin.

There's a slight difference between the last two circuits in that Hot Cake sets clip points off the base of the transistors.  Whereas the diode case has the threshold resistors are part of the feedback network.   The supply divider adds series resistance to the diodes.    Either scheme works for diodes and transistors.  The main thing about the transistor case is the amount of effective series resistance can be much reduced.    Using low valued resistor in the diode case isn't practical, one thing you can do through is wire the diodes to a zener or a voltage with a large cap.

These are all old ideas perhaps going back to the 60's.


EDIT:
Here's a transistor version.  More or less the feedback version of the Hot Cake in the sense that the thresholds are programmed off the supply rails.     This one has two threshold so you can simplify by using only two transistors.    You can also remove the diode part.  The resistors in series with the emitters can also be removed for harder clipping (the part with the trimpots).